(sorry to abandon this thread - I went away for the weekend)
I have indeed been in a classroom, and even had some teaching experience, although I don't claim to be an education expert.
I have 2 problems with your approach. First, I don't want to over-rely on the reports of smart teachers. Not because I don't trust them, but because when it comes to public education we have limited financial resources, and a distribution of ability across the population of teachers. Maybe a smart teacher can consistently get better results with a smaller class. I could believe it. Unfortunately, we have yet to find a system that ensures we only hire and retain smart teachers. In matters of public policy, we are often forced to forgo optimal results for some, in order to achieve acceptable results for the system as a whole.
My second problem is more fundamental: how are we evaluating what's 'better'? If this determination is made purely by those who are most involved, then logic dictates that parents should be the ultimate arbiters of educational quality by virtue of having the greatest intimacy with their kids. Of course, we know that this is not the case in practice, because parents' commitment and ability varies very widely indeed, and so we cannot trust them as objective reporters. If we did, collage admissions officers and potential employers could simply call up an applicant's parents and ask them if the applicant would be a good candidate. In reality, we use testing of various kinds because it gives us some sort of objective yardstick with which to measure academic performance across a large population. We simply don't have the resources to fully evaluate every aspect of every individual candidate, and while statistical methods are rather dehumanizing they do offer enough predictive power to be useful.
Consider that while High School teachers express a high degree of optimism about the future academic prospects of their graduates, college teachers have been complaining more and more about the number of incoming students who need remedial classes in English or Math. Just as an excess of 'teaching to the test' may measure only test-taking ability, an excess reliance on classroom performance as reported by teachers may measure only the ability to perform in a classroom context.